Diagnosis of system health with event logs

1. A method for diagnosing system health with system event logs, the method comprising: receiving a plurality of text event logs and health indicator states from a system; transducing the plurality of text event logs into numeric-based metrics of the system, wherein transducing includes organizing textual event messages of the plurality of text event logs into clusters based on a similarity between the textual event messages and based on whether the similarity is greater than a similarity threshold; and deriving, based on the transduced numeric-based metrics, at least one model of the system that correlates the plurality of event logs to the corresponding health indicator states.

2. The method of claim 1 , wherein transducing comprises: transducing the textual event messages in the plurality of text event logs into the numeric-based metrics indicating one or more conditions of the system or at least one application executing therein.

3. The method of claim 1 , wherein transducing comprises: computing one of the numeric-based metrics for each of the clusters based on the organized clusters.

4. The method of claim 3 , wherein the step of organizing comprises: providing a distance function for clustering the plurality of text event logs; providing the similarity threshold; providing a cluster set; computing a distance between each of the textual event messages and each cluster found in the cluster set, based on the provided distance function; comparing the computed distance with the provided similarity threshold; responsive to the computed distances being smaller than the provided similarity threshold, adding the textual event message associated with the computed distance as a new cluster in the cluster set.

5. The method of claim 4 , wherein the step of organizing further comprises: responsive the computed distance being greater than or equal to the provided similarity threshold, adding a count to the cluster associated with the computed distance.

6. The method of claim 5 , wherein the step of computing comprises: computing a value for each one of the numeric-based metrics for each cluster found in the cluster set by aggregating a number of counts in the each cluster.

7. The method of claim 6 , wherein the step of deriving comprises: deriving the at least one system model correlating each of the computed numeric-based metrics with one of the health indicator states.

8. The method of claim 4 , wherein the step of providing the cluster set comprises: initializing the cluster set as an empty set.

9. The method of claim 1 , wherein the step of receiving comprises: receiving the plurality of text event logs and health indicator states over a plurality predefined time periods, wherein there is at least one of the plurality of text event logs and one of the health indicator states corresponding to each of the plurality of predefined time periods.

10. The method of claim 4 , wherein the distance function is a cosine distance function for determining a similarity between two textual event messages.

11. The method of claim 1 , wherein the step of transducing comprises: distilling, from the organizing of textual event messages, a set of prototypical event messages from the plurality of text event logs; counting a number of times each of prototypical event messages appears in a predefined time period; and setting the count for each of the prototypical event messages as one of the transduced numeric-based metrics.

12. A system for providing automated health diagnosis of a computing system, comprising: a metrics transducer module operates to receive a plurality of text event logs and health indicator states of the computing system and to transduce the plurality of text event logs into numeric-based metrics of the system, wherein the metrics transducer module organizes textual event messages of the plurality of text event logs into clusters based on a similarity between the textual event messages and based on whether the similarity is greater than a similarity threshold; and a model building engine, executed by a processor, operates to derive, based on the transduced numeric-based metrics, at least one model of the system that correlates the plurality of text event logs to the corresponding health indicator states.

13. The system of claim 12 , wherein: each of the transduced numeric-based metrics includes an identification of the each transduced numeric-based metric and a numerical value for the each transduced numeric-based metric.

14. The system of claim 13 , wherein the numerical value for one of the transduced numeric-based metrics indicates a number of the plurality of textual event messages that is categorized into a defined group of messages.

15. The system of claim 14 , wherein the metrics transducer module includes a distance function that operates to categorize the number of the plurality of textual event messages into the defined group of messages.

16. The system of claim 15 , wherein the distance function is a cosine distance function for determining a similarity between two text messages.

17. The system of claim 12 , wherein the metrics transducer module further computes a distance between each of the textual event messages and each of the clusters based on a distance function, compares the computed distance with the similarity threshold, and, if the computed distance is smaller than the similarity threshold, adds the textual event message associated with the computed distance as a new cluster into the clusters.

18. The system of claim 17 , wherein the clusters are dynamically defined by the textual event messages in the plurality of text event logs.

19. A computer readable non-transitory medium on which is encoded computer executable programming code that includes computer execution instructions to: receive a plurality of text event logs and health indicator states from a system; transduce the plurality of text event logs into numeric-based metrics of the system including organize textual event messages of the plurality of text event logs into clusters based on a similarity between the textual event messages and based on whether the similarity is greater than a similarity threshold; and derive, based on the transduced numeric-based metrics, at least one model of the system that correlates the plurality of text event logs to the corresponding health indicator states.

20. The computer-readable non-transitory medium of claim 19 , wherein the numeric-based metrics indicating indicate one or more conditions of the system or at least one application executing therein.